forecasting purposes in a (pseudo) real-time setting, when the current production

data as well as their immediate lags would not have been published yet (often called


2 Data and empirical approach

The dependent variables that we analyze are the growth rate of German real total

industrial production (IP) and the production in the construction sector. Total output

represents an important cyclical indicator, while production in the construction

sector is the part of economic activity which is most likely to depend on weather

conditions. An overview about the different production indices and their hierarchial

structure is given in Statistisches Bundesamt (2015). Data are taken from the

Bundesbank website. Both indices are calendar and seasonally adjusted.

Weather data for Germany have recently begun to be provided on a daily basis

and are freely available. 1 The construction of the weather dataset was inspired by

the approach of Hummel, Vosseler, Weber, and Weigand (2015), that is we aggregated

daily weather data of the 251 weather stations available to the federal state

level, then weighing them by the state-level number of employees to obtain aggregated

data at the national level. The sample used in this paper is January 1991

through October 2015. We consider three measurable weather aspects, namely air

temperature, snow height in cm, and snow fall per week in cm, all time-averaged

from daily to monthly series. Other weather variables would also be possible in

principle; the Deutsche Bundesbank (2014) for example used the sum of ice-days

in a specific time interval (quarter or month), but that information should not differ

much from the combined content of snow fall and (cumulated) height.

Given that the weather data are published almost immediately – in contrast to

1 Original database provided by Deutscher Wetterdienst and freely available at http://www.


More magazines by this user
Similar magazines